Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Wed, 16 Jun 2021 16:07:01 +0300
From:   Vladimir Oltean <olteanv@gmail.com>
To:     Vadym Kochan <vadym.kochan@plvision.eu>
Cc:     "David S. Miller" <davem@davemloft.net>,
        Jakub Kicinski <kuba@kernel.org>, netdev@vger.kernel.org,
        Andrew Lunn <andrew@lunn.ch>,
        Taras Chornyi <tchornyi@marvell.com>,
        linux-kernel@vger.kernel.org,
        Mickey Rachamim <mickeyr@marvell.com>,
        Serhiy Boiko <serhiy.boiko@plvision.eu>,
        Volodymyr Mytnyk <vmytnyk@marvell.com>,
        Vadym Kochan <vkochan@marvell.com>
Subject: Re: [PATCH net-next 1/2] net: marvell: Implement TC flower offload
Message-ID: <20210616130701.e4k2izlthmj5j6yc@skbuf>
References: <20210615125444.31538-1-vadym.kochan@plvision.eu>
 <20210615125444.31538-2-vadym.kochan@plvision.eu>
 <20210616005453.cuu3ocedgfcafa7o@skbuf>
 <20210616130424.GB9951@plvision.eu>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20210616130424.GB9951@plvision.eu>
Precedence: bulk

On Wed, Jun 16, 2021 at 04:04:24PM +0300, Vadym Kochan wrote:
> Hi Vladimir,
> 
> On Wed, Jun 16, 2021 at 03:54:53AM +0300, Vladimir Oltean wrote:
> > On Tue, Jun 15, 2021 at 03:54:43PM +0300, Vadym Kochan wrote:
> > > +static int prestera_port_set_features(struct net_device *dev,
> > > +				      netdev_features_t features)
> > > +{
> > > +	netdev_features_t oper_features = dev->features;
> > > +	int err;
> > > +
> > > +	err = prestera_port_handle_feature(dev, features, NETIF_F_HW_TC,
> > > +					   prestera_port_feature_hw_tc);
> > 
> > Why do you even make NETIF_F_HW_TC able to be toggled and not just fixed
> > to "on" in dev->features? If I understand correctly, you could then delete
> > a bunch of refcounting code whose only purpose is to allow that feature
> > to be disabled per port.
> > 
> 
> The only case where it can be used is when user want to disable TC
> offloading and apply set of rules w/o skip_hw.
> 
> So you think it is OK to not having an ability to disable offloading at
> all ?

Because adding "skip_hw" is already possible per filter in the first
place, I don't think that this feature justifies the added complexity, no.